WorldClim Version2 uses data from 1970-2000

[Alexandre Silva de Paula]

Dear all!

I would d like to know what time interval is most appropriate for potential distribution analysis. WorldClim Version2 uses data from 1970-2000. I have been reading articles that use 1910 collection data, for example. Is that correct? Could someone clarify me if the correct one would be to use only data collected between 1970-2000? Could I include more recent data, post 1970-2000?

Best wishes,

Alexandre Silva de Paula.

[Adam Smith]

Hi Alexandre,
That's a good question... Usually you want to use collections and environmental data that overlap in time, but it's not uncommon to use collections somewhat outside of a climate dataset's time period. For example, as you note, WC covers on your up to 2000 and CHELSA to 2013, but if you have more recent collections data, what are you going to do? In my mind there are two things to consider.

First, consider that WC, CHELSA, ClimateNA, ClimateEU, ClimateLA, CRU, and the many other climate data sources used by SDMers often represent "normals" or averages across 30-yr periods. This is a standard time period used to remove transient events that vary across years or decades like El Nino or the Pacific Decadal Oscillation. So in a sense the normals represent no particular year even in the time period across which they're averaged. In light of that it may not be too egregious to use collections data outside the normal's time period so long as you expect the year(s) in which the collections were made to be similar to the years inside the time period (ie, it could have come from the same distribution, even if it does not match the climate in any particular year of that period).

Given that, consider if there has been a major change in trend in climate between the time period of the climate data and the collections. For example, warming really started to accelerate in western North America in the 1970s. And globally the hottest 20 years on record all came since 1995. So maybe it's not a good idea to use very recent collections data with a 20th century climate data. But, I know that in saying this in many cases it's the only climate data available for many places. So there's little alternative.

To my knowledge this issue been looked at, though people have compared using climate versus weather as predictors (which is a related question).

Anyone else have insights?

Adam